Timo Oess: From Sounds to Locations - Computational models of the early auditory pathway

Abstract

Incoming sound waves produce small vibrations of different frequencies at the eardrums of the left and right ear. By ordering these vibrations into different frequency channels the cochlea forms a tonotopical organization of sounds. Animals, including humans, require intensive processing steps along the auditory pathway to perceive, localize and identify sounds from these neural activations. To extract spatial locations of sounds, the auditory system integrates signals from both ears for computation of interaural differences in arrival time (ITD) and perceived sound level (ILD). Such cues enable the localization of sounds in the horizontal plane but create ambiguous information in the vertical plane. Thus, elevation-dependent spectral cues created by the shape of the ear are analyzed to localize sounds in the vertical plane. Together, these cues are evaluated to generate a 3D representation of auditory space. In order to form a common frame of reference and a means for multisensory integration, this space needs to be calibrated by visual position signals. Computational models that describe such processing steps are currently either missing or merely specialize on a single stage of the auditory pathway.

In this talk, I investigate the auditory pathway for localizing sound sources by developing canonical computational neural network models of consecutive processing stages. In particular, I present a model for ILD computation in the lateral superior olive incorporating a novel synaptic adaptation mechanism that leads to improved spatial localization of sounds in the horizontal plane. Furthermore, I introduce a novel concept of how spectral cues could be processed for vertical sound localization in monaural and binaural conditions. Results suggest that sound source localization in the vertical plane is fundamentally binaural in contrast to common assumptions. In a model of the inferior colliculus, I present how these cues are aligned with a visual teacher signal to provide a common frame of reference. Finally, a multisensory integration model of audio-visual inputs of the superior colliculus demonstrates the importance of cortical feedback connections and provides insights on its effect for Bayesian optimal fusion. Model conception and simulation experiments investigate underlying neural principles and provide new hypotheses for auditory sound source localization.

 About the speaker and his research
Timo Oess

Hosted by Enya Paschen